On Friday 25 March 2011 20:47:22 Tony Travis wrote: > OK, how about attempting to assemble the smaller 'large' insert > libraries with MIRA, and then mapping the 500bp reads onto that? > 44x 2 x 114nt paired-end --> 500bp <-- > 8x 2 x 36nt mate-pair <-- 3kb --> > 8x 2 x 36nt mate-pair <-- 5kb --> > 4x 2 x 36nt mate-pair <-- 10kb --> > Just a thought... De-novo assemblies of reads <75bp is problematic, and <50bp is just plain evil. At least when one expects reasonable results. 75bp starts to be nice, and >100bp is "fun" again. Then there's another problem: just one of the 8x coverage libraries (which would be by far not enough anyway) of 1.8 gb with 36bp reads represents ~450m reads! Not really much less than 800m reads. If I made my calculations right, the whole data set has something like 1.8 to 2 billion reads. B.